mcpField
O
orchestrate-multi-target-sdks

orchestrate-multi-target-sdks

github.com/speakeasy-apiDevelopment1888 installsUpdated 37w ago
Repository

About orchestrate-multi-target-sdks

A skill for orchestrating SDK generation across languages or API variants from a repository with the Speakeasy CLI. It configures sources and targets through commands, then generates all or selected SDKs.

Key features

  1. 01Configure OpenAPI sources by CLI
  2. 02Add language-specific SDK targets
  3. 03Generate SDKs for multiple API variants
  4. 04Run generation for all or selected targets
  5. 05Set up SDK monorepo workflows

Use cases

  1. 01Generate TypeScript, Python, and Go SDKs from one spec
  2. 02Create SDK variants from separate OpenAPI specs
  3. 03Add a language target to an existing workflow
  4. 04Run SDK generation for selected targets

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/speakeasy-api/sdk-generation-skills --skill orchestrate-multi-target-sdks

After installing, type/orchestrate-multi-target-sdksin Claude Code to confirm it loaded.

Details

From the project README

Orchestrate Multi-Target SDKs

Generate SDKs for multiple languages or variants from a single repository using CLI commands.

When to Use

  • Generating SDKs for multiple languages from the same API spec
  • Creating SDK variants (Azure, GCP, regional) from different specs
  • Setting up an SDK monorepo
  • User says: "generate SDKs for multiple languages", "SDK for each language", "multi-target"

Quick Start: Multiple Languages

Always use CLI commands. Never create .speakeasy directories manually.

# Step 1: Initialize first target (creates .speakeasy/workflow.yaml)
speakeasy quickstart --skip-interactive --output console \
  -s openapi.yaml -t typescript -n "MySDK" -p "my-sdk"

# Step 2: Add more language targets using configure
speakeasy configure targets \
  --target-type python \
  --source my-source \
  --output ./sdks/python

speakeasy configure targets \
  --target-type go \
  --source my-source \
  --output ./sdks/go

# Step 3: Generate all SDKs
speakeasy run --output console

CLI Reference

speakeasy configure sources

Add a new OpenAPI source to an existing workflow:

speakeasy configure sources \
  --location ./openapi.yaml \
  --source-name my-api

# With authentication header
speakeasy configure sources \
  --location https://api.example.com/openapi.yaml \
  --source-name my-api \
  --auth-header "Authorization"
FlagShortDescription
--location-lOpenAPI document location (file or URL)
--source-name-sName for the source
--auth-headerAuthentication header name (optional)
--output-oOutput path for compiled source (optional)
--non-interactiveForce non-interactive mode
speakeasy configure targets

Add a new SDK target to an existing workflow:

speakeasy configure targets \
  --target-type typescript \
  --source my-api \
  --output ./sdks/typescript

# With all options
speakeasy configure targets \
  --target-type go \
  --source my-api \
  --target-name my-go-sdk \
  --sdk-class-name MyAPI \
  --package-name github.com/myorg/myapi-go \
  --output ./sdks/go
FlagShortDescription
--target-type-tLanguage: typescript, python, go, java, csharp, php, ruby, terraform
--source-sName of source to generate from
--target-nameName for the target (defaults to target-type)
--sdk-class-nameSDK class name (optional)
--package-namePackage name (optional)
--base-server-urlBase server URL (optional)
--output-oOutput directory (optional)
--non-interactiveForce non-interactive mode

Example: Multi-Language SDKs

Single OpenAPI spec → TypeScript, Python, Go SDKs:

# Initialize with first target
speakeasy quickstart --skip-interactive --output console \
  -s ./openapi.yaml -t typescript -n "MySDK" -p "my-sdk" -o ./sdks/typescript

# Add Python
speakeasy configure targets \
  --target-type python \
  --source my-source \
  --sdk-class-name MySDK \
  --package-name my-sdk \
  --output ./sdks/python

# Add Go
speakeasy configure targets \
  --target-type go \
  --source my-source \
  --sdk-class-name MySDK \
  --package-name github.com/myorg/my-sdk-go \
  --output ./sdks/go

# Generate all
speakeasy run --output console

Example: SDK Variants (Multiple Sources)

Different OpenAPI specs → variant SDKs:

# Initialize with main API
speakeasy quickstart --skip-interactive --output console \
  -s ./openapi.yaml -t typescript -n "MySDK" -p "my-sdk"

# Add Azure variant source
speakeasy configure sources \
  --location ./openapi-azure.yaml \
  --source-name azure-api

# Add Azure target
speakeasy configure targets \
  --target-type typescript \
  --source azure-api \
  --target-name typescript-azure \
  --sdk-class-name MySDKAzure \
  --package-name "@myorg/my-sdk-azure" \
  --output ./packages/azure

# Generate all
speakeasy run --output console

Repository Structure

my-api-sdks/
├── openapi.yaml              # Source spec
├── .speakeasy/
│   └── workflow.yaml         # Multi-target config (created by CLI)
├── sdks/
│   ├── typescript/
│   │   ├── .speakeasy/
│   │   │   └── gen.yaml      # Created by configure
│   │   └── src/
│   ├── python/
│   │   ├── .speakeasy/
│   │   │   └── gen.yaml
│   │   └── src/
│   └── go/
│       ├── .speakeasy/
│       │   └── gen.yaml
│       └── go.mod
└── .github/workflows/
    └── sdk_generation.yaml

Running Generation

# Generate all targets
speakeasy run --output console

# Generate specific target only
speakeasy run -t typescript --output console
speakeasy run -t python --output console

CI Workflow

# .github/workflows/sdk_generation.yaml
name: Generate SDKs
on:
  push:
    branches: [main]
    paths: ['openapi.yaml']
  workflow_dispatch:

jobs:
  generate:
    uses: speakeasy-api/sdk-generation-action/.github/workflows/workflow-executor.yaml@v15
    with:
      mode: pr
    secrets:
      github_access_token: ${{ secrets.GITHUB_TOKEN }}
      speakeasy_api_key: ${{ secrets.SPEAKEASY_API_KEY }}

What NOT to Do

  • Do NOT create .speakeasy/ directories manually - Use speakeasy quickstart and speakeasy configure
  • Do NOT write workflow.yaml or gen.yaml files directly - Use CLI commands
  • Do NOT copy .speakeasy/ directories between projects - Each needs its own config
Incorrect
# WRONG: Do not do this
mkdir -p .speakeasy
cat > .speakeasy/workflow.yaml << 'EOF'
...
EOF
Correct
# RIGHT: Use CLI commands
speakeasy quickstart --skip-interactive --output console \
  -s openapi.yaml -t typescript -n "MySDK" -p "my-sdk"

speakeasy configure targets --target-type python --source my-source

Troubleshooting

IssueSolution
Wrong target generatedSpecify -t target-name in speakeasy run
Source not foundRun speakeasy configure sources to add it
Target not foundRun speakeasy configure targets to add it
Config out of syncRun speakeasy run to regenerate

After Making Changes

After adding sources or targets, regenerate:

speakeasy run --output console

After Making Changes

After modifying workflow.yaml or per-target gen.yaml, prompt the user to regenerate the SDK(s):

Configuration complete. Would you like to regenerate the SDK(s) now with speakeasy run?

If the user confirms, run:

# Generate all targets
speakeasy run --output console

# Or generate a specific target
speakeasy run -t <target-name> --output console

Changes to workflow.yaml and gen.yaml only take effect after regeneration.

Related Skills

  • start-new-sdk-project - Initial SDK setup for single target
  • configure-sdk-options - Language-specific gen.yaml options
  • manage-openapi-overlays - Spec customization with overlays
  • orchestrate-multi-repo-sdks - Separate repository per language

Comments

Markdown supported